Improved method for detecting external fire causes in mines using YOLOv9m

doi:10.16265/j.cnki.issn1003-3033.2026.06.0726

Abstract

Abstract:

To address the problem of strong background interference, slow detection speed, and a high missed-detection rate in the detection of exogenous mine fires, an improved YOLOv9m-based method for object detection and image recognition was proposed. Firstly, the convolution modules in the backbone network of the original model were replaced with the lightweight paddlepaddle lightweight convolutional network (PP-LCNet) module, which reduced the number of model parameters. Secondly, the fully contextual attention (FCAttention) module was embedded into the backbone network to enhance flame feature interaction and optimize weight allocation, thereby improving the accuracy of feature selection and information fusion in complex scenes. Finally, the content-ware reassembly of features (CARAFE) dynamic upsampling operator was introduced into the neck network. Through a content-aware mechanism, feature information was dynamically reorganized, and the accuracy of detail representation was improved. The results show that, compared with the original model, the number of parameters and floating-point operations of the improved model are reduced by 13.2% and 13.6%, respectively, while precision, mean average precision (mAP), and frames per second(FPS) are increased by 5.2%, 3.2%, and 23.2%, respectively. The improved YOLOv9m algorithm ensures real-time fire detection accuracy in complex underground environments. It also significantly improves the detection accuracy in small-target fire-source scenarios. The proposed method meets the requirements of lightweight design and real-time performance, and it provides support for early fire warning and rapid emergency response.

Key words: external mine fire, YOLOv9m, object detection, lightweight, attention mechanism

CLC Number:

X932爆炸安全与防火、防爆

Qi Yun, Yao Rui, Zhan Xinhui, Xue Kailong, Jing Xueyan, Qi Qingjie. Improved method for detecting external fire causes in mines using YOLOv9m[J]. China Safety Science Journal, 2026, 36(6): 110-118.

Figures/Tables 14

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Table 1

Fig.6

Fig.7

Fig.8

Fig.9

Table 2

Table 3

Fig.10

Fig.11

References 22

[1]	Kodur V, Kumar P, Rafi M M. Fire hazard in buildings: review, assessment and strategies for improving fire safety[J]. PSU Research Review, 2020, 4(1):1-23. doi: 10.1108/PRR-12-2018-0033
[2]	王德明, 程远平, 周福宝, 等. 矿井火灾火源燃烧特性的实验研究[J]. 中国矿业大学学报, 2002, 31(1): 33-36.
	Wang Deming, Cheng Yuanping, Zhou Fubao, et al. Experimental research on combustion property of mine fire source[J]. Journal of China University of Mining &Technology, 2002, 31(1): 33-36.
[3]	朱红青, 胡超, 张永斌, 等. 我国矿井内因火灾防治技术研究现状[J]. 煤矿安全, 2020, 51(3): 88-92.
	Zhu Hongqing, Hu Chao, Zhang Yongbin, et al. Research status on prevention and control technology of coal spontaneous fire in China[J]. Safety in Coal Mines, 2020, 51(3): 88-92.
[4]	李海, 熊升华, 孙鹏. 基于特征工程的S-FCN火灾图像检测算法[J]. 中国安全科学学报, 2024, 34(9):191-201. doi: 10.16265/j.cnki.issn1003-3033.2024.09.2063
	Li Hai, Xiong Shenghua, Sun Peng. S-FCN fire image detection method based on feature engineering[J]. China Safety Science Journal, 2024, 34(9):191-201.
[5]	汤伟, 张文迪, 袁航, 等. 基于YOLOv7的红外阴燃火探测算法改进研究[J]. 燃烧科学与技术, 2024, 30(5):532-538.
	Tang Wei, Zhang Wendi, Yuan Hang, et al. Improvement of infrared smoldering fire detection algorithm based on YOLOv7[J]. Journal of Combustion Science and Technology, 2024, 30(5):532-538.
[6]	Ren Shaoqing, He Kaiming, Girshick R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 39(6): 1137-1149. doi: 10.1109/TPAMI.2016.2577031
[7]	Dai Jinfeng, Li Yi, He Kaiming, et al. R-FCN: object detection via region-based fully convolutional networks[J]. Advances in Neural Information Processing Systems, 2016, 29: DOI:10.48550/arXiv.1605.06409.
[8]	He Kaiming, Gkioxari G, Dollár P, et al. Mask R-CNN[C]. Proceedings of the IEEE International Conference on Computer Vision, 2017:2961-2969.
[9]	Redmon J, Divvala S, Girshick R, et al. You only look once: unified, real-time object detection[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016:779-788.
[10]	张帆, 张嘉荣, 程海星. 基于深度学习的矿井智能目标检测技术研究综述[J]. 煤炭科学技术, 2025, 53(增刊1): 284-296.
	Zhang Fan, Zhang Jiarong, Cheng Haixing. Research review on intelligent object detection technology for coal minesbased on deep learning[J]. Coal Science and Technology, 2025, 53(S1):284-296.
[11]	Liu Yin, Wen Hu, Chen Changming, et al. Research status and development trend of coal spontaneous combustion fire and prevention technology in China: a review[J]. Acs Omega, 2024, 9(20): 21 727-21 750.
[12]	秦波涛, 仲晓星, 王德明, 等. 煤自燃过程特性及防治技术研究进展[J]. 煤炭科学技术, 2021, 49(1): 66-99.
	Qin Botao, Zhong Xiaoxing, Wang Deming, et al. Research progress of coal spontaneous combustion process characteristics and prevention technology[J]. Coal Science and Technology, 2021, 49(1): 66-99.
[13]	胡纪年, 李雨成, 李俊桥, 等. 基于CNN的矿井外因火灾火源定位方法研究[J]. 中国安全生产科学技术, 2024, 20(3):134-140.
	Hu Jinian, Li Yucheng, Li Junqiao, et al. Study on localization method of mine exogenous fire source based on CNN[J]. Journal of Safety Science and Technology, 2024, 20(3):134-140.
[14]	Diwan T, Anirudh G, Tembhurne J V. Object detection using YOLO: challenges, architectural successors, datasets and applications[J]. Multimedia Tools and Applications, 2023, 82(6):9243-9275. doi: 10.1007/s11042-022-13644-y
[15]	Chien W, Yeh I, Mark L. YOLOV9: learning what you want to learn using programmable gradient information[C]. European Conference on Computer Vision, 2024: 1-21.
[16]	Ye Zhihui, Wu Jian, Zhao Xiaozhong, et al. Multimodal object detection based on feature interaction and adaptive grouping fusion[J]. Infrared Technology, 2025, 47(4): 468-474.
[17]	Sun Hang, Wen Yang, Feng Huijing, et al. Unsupervised bidirectional contrastive reconstruction and adaptive fine-grained channel attention networks for image dehazing[J]. Neural Networks, 2024, 176:DOI:10.1016/j.neunet.2024.106314.
[18]	Wang Jiaqi, Chen Kai, Xu Rui, et al. CARAFE: content-aware reassembly of features[J]. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2019: 3007-3016.
[19]	Gao Jinfeng, Chen Yu, Wei Yongming, et al. Detection of specific building in remote sensing images using a novel YOLO-S-CIOU model. case: gas station identification[J]. Sensors, 2021, 21(4): DOI:10.3390/s21041375.
[20]	陈晓明, 张雅丽. 基于改进YOLOv8n的夜间行人跌倒检测算法:CCFM-YOLOv8n[J]. 科学技术与工程, 2025, 25(35):15174-15 183.
	Chen Xiaoming, Zhang Yali. Pedestrian fall detection algorithm at night based on improved YOLOv8n: CCFM-YOLOv8n[J]. Science Technology and Engineering, 2025, 25(35): 15 174-15 183.
[21]	祁云, 薛凯隆, 汪伟, 等. 矿井煤与瓦斯突出事故应急救援能力评估模型[J]. 中国安全科学学报, 2024, 34(2):225-230. doi: 10.16265/j.cnki.issn1003-3033.2024.02.0983
	Qi Yun, Xue Kailong, Wang Wei, et al. Assessment model of emergency response capability for coal and gas outburst accidents in mines[J]. China Safety Science Journal, 2024, 34(2):225-230. doi: 10.16265/j.cnki.issn1003-3033.2024.02.0983
[22]	蒋仕新, 邹小雪, 杨建喜, 等. 复杂背景下基于改进YOLO v8s的混凝土桥梁裂缝检测算法[J]. 交通运输工程学报, 2024, 24(6):135-147.
	Jiang Shixin, Zou Xiaoxue, Yang Jianxi, et al. Concrete bridge crack detection method based on improved YOLOv8s in complex backgrounds[J]. Journal of Traffic and Transportation Engineering, 2024, 24(6):135-147.

参数	数值
图像尺寸/px×px×px	640×640×3
学习率	0.001
动量系数	0.937
批量处理	8
训练周期/轮	300
权重衰减系数	0.005

PP-LCNet	FCAttention	CARAFE	MPDIoU	P/ %	R/ %	mAP@0.5/ %	参数量/ MB	运算次数/ GB	检测速度/ (帧/s)
×	×	×	×	92.59	87.36	93.49	32.55	130.72	29.50
√	×	×	×	83.79	76.04	84.72	19.53	83.901	59.49
√	√	×	×	91.20	79.61	89.00	27.12	104.90	27.79
√	√	√	×	94.32	80.43	95.46	28.12	112.97	35.35
√	√	√	√	97.36	84.91	96.49	28.12	112.97	36.33

模型	P/%	R/%	mAP@0.5/%	参数量/MB	运算量/GB
Fast R-CNN	81.62	80.86	83.76	128.55	247.76
YOLOv5	94.07	82.77	94.91	34.63	126.78
YOLOv7	92.48	81.34	91.42	35.79	136.67
YOLOv8	92.59	87.36	94.49	32.45	140.72
PPL-YOLOv9m-F-C	97.36	84.91	96.49	28.12	112.97